‘Whose Language Counts?’ Linguistic Racism, Gen‐AI, and Policy Silence in Polish and Romanian AI Governance

Abstract

ABSTRACT As generative artificial intelligence spreads through European education, its development continues to be shaped overwhelmingly by English and other dominant languages. This article examines how national AI and digital education policies in Poland and Romania discursively construct a linguistically homogeneous ‘ideal learner,’ rendering Roma and refugee learners invisible in policy. Drawing on comparative critical discourse analysis of four national policy documents alongside three civil society reports, the analysis identifies three patterns: technocratic and economic nationalist discourse that presupposes linguistic homogeneity; rhetorical commitments to inclusion that erase linguistic diversity in practice; and the complete absence of guidance on linguistic variation in teacher preparation. Situating these dynamics within post‐socialist legacies of centralized linguistic governance, the article argues that AI governance operates as implicit language policy, and that treating it as such is a necessary condition for contesting the linguistic hierarchies these systems reproduce.